A

Aodttca©

K

1

sblfatmeomv

ibopa

0d

Available online at www.sciencedirect.com

Journal of the European Ceramic Society 30 (2010) 3187–3201

Friction surface evolution of carbon fibre reinforced carbon/silicon carbide(Cf/C-SiC) composites

Yuan Wang, Houzheng Wu ∗Department of Materials, Loughborough University, Leicestershire LE11 3TU, UK

Received 10 March 2010; received in revised form 24 June 2010; accepted 9 July 2010Available online 4 August 2010

bstract

study is reported on the development of friction surface of carbon fibre reinforced ceramic composites through microstructural image registrationf the surface after a range of braking stops on a laboratory-scale dynamometer test rig. It has been found that a steady friction transfer layer iseveloped in silicon regions; in carbon fibre/carbon and most silicon carbide regions, the friction surface is unsteady and any possible frictionransfer layer is hardly built up with satisfactory longevity. Large voids and cracks/crevices are likely to be filled with transferred materials, but

hese compacts are susceptible to be stripped off by further braking operation. From this study, three types of friction surfaces are identified andould exist in bedding stage. Under the current testing configuration and regime, coefficient of friction increases with accumulated braking stops,nd no stable friction yet appeared. The development of friction surface and its impact on the friction measurements are discussed.

2010 Elsevier Ltd. All rights reserved.

letafuaCrtnCtt

en

eywords: A. Composite; B. Fibres; C. SiC; D. Wear parts; E. Friction

. Introduction

Carbon fibre reinforced ceramic composites have progres-ively been used as discs, or rotors as some call them, of frictionrakes in transport vehicles to benefit from the light weight andong durability that the combination of carbon fibres (as rein-orcements) and ceramics (as matrix) can provide when they areppropriately engineered together through tailoring microstruc-ure of the composites.1 Among these composites, the one with a

atrix consisting of carbon and silicon carbide has been mostlyxploited; in the real world, silicon could be another constituentf the matrix particularly when silicon infiltration is included inanufacture, whilst this type of composite is frequently abbre-

iated as Cf/C-SiC.2–6

Friction performance of Cf/C-SiC composites has been stud-ed under various braking contexts. It is noticed that the frictionehaviour of these composites is in general different from that

f metallic materials, e.g. grey cast irons and metal matrix com-osites, when they are tested against a counterpart, typicallyn organic or sintered-metallic pad. Such a distinction could

∗ Corresponding author. Tel.: +44 1509 223342; fax: +44 1509 223949.E-mail address: h.wu2@lboro.ac.uk (H. Wu).

ffCmast

955-2219/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.oi:10.1016/j.jeurceramsoc.2010.07.019

ead to particular difficulties in implementing knowledge andxperiences that have been accumulated for a long period ofime in the development of pad materials for metallic discs;

consequence of this could create a potential impediment inulfilling the potentials that Cf/C-SiC can provide. A furthernderstanding is therefore needed to underpin possible mech-nisms that could govern the observed friction behaviour off/C-SiC composites when they are engaged with pad mate-

ials. Such an understanding is likely to provide guidelines forhe development of pad materials, and more importantly rightow, for any further attempts to optimize the microstructure off/C-SiC composites or their friction bearing surfaces, in order

o achieve the required friction and engineering performance ofhe friction brakes.

Friction performance of Cf/C-SiC composites is normallyxamined under a disc to pad or disc to disc configuration. It isoticed that the microstructural constituents of the friction sur-ace of the composite disc could have a strong influence on theriction performance and wear resistance. For instance, when the/Cf-SiC discs were coupled with carefully selected sintered-

etallic pads, a higher, more stable coefficient of friction (COF),

nd a lower wear rate were achieved for the couple with a frictionurface of the composite consisting of silicon carbide only thanhat with a surface containing more than just silicon carbide, such

3 ean C

asmtae

amwtoPncramwpttr(tcmb

cbkcp(lctfaalfmaoilIawodcsFc

tfm

oiTtidcpg

2

2

ivpctTdawft

sgm(wgpa

ssaaf

2

ddvaIn each braking test, the Cf/C-SiC disc rotates, with its kinetic

188 Y. Wang, H. Wu / Journal of the Europ

s carbon, carbon fiber filaments, and silicon.4 This observationignifies that the inherent microstructure of these compositesay provide a complexity of friction surface to accommodate

he braking context; some microstructural constituents and char-cteristics on the surface may have less synergy or even negativeffects on the performance of a brake.

It was also noticed that, when such a composite was set togainst itself, i.e. disc to disc configuration, the friction perfor-ance, including the level of COF and its stability, was improvedith the reduction of porosity and the increase of carbon con-

ent, and the wear rate became more dependent on the resistancef silicon carbide regions on the thermo/mechanical stresses.7

ost-mortem examination on the friction surface indicated thato significant friction transfer layer/film was developed, exceptrumbled carbon and carbon fibre debris. Afterward, sameesearchers believed that the graphitization of the carbon and theppearance of friction transferred layer on the friction surfaceight have benefited the improvement in friction performance,8

hilst a more ad hoc analysis is needed to unveil the under-inning reasons at a microstructural level. It was also claimedhat Cf/C-SiC composites manufactured by liquid silicon infil-ration (LSI) had a higher coefficient of friction and lower wearate than those manufactured by chemical vapour infiltrationCVI).3 These experimental observations seem to invoke a needhat a further understanding of the friction performance of theseeramic composites should concentrate on what happens at theicrostructural level, apart from friction performance evaluation

ased on engineering tests.It has been noticed that the counterpart of the composite disc

ould play a key role as well in determining the level and the sta-ility of coefficient of friction for a specified friction couple. Twoinds of materials have been developed for pads to against theeramic composite discs, i.e. polymeric material bonded com-osite (often called organic) and sintered-metallic compositecalled sintered). The formulation of the pad materials couldead to a wide range of friction performance by against the sameeramic disc. For instance, when sintered pads were slightlywisted in its formulation, the coefficient of friction could varyrom 0.3 for bronze pad with silicate reinforcements, to 0.65 withfurther addition of graphite and increase of silicate9; Stadler etl. also noticed that softening the sintered pad by adding graphiteed to an increase of COF.10 In the development of pad materialsor metallic rotors, one of the major focuses is on the develop-ent of friction transfer films on the friction surfaces of both pad

nd disc. It is believed that a reaction product could be generatedn the mating surfaces of the friction couple during friction brak-ng; it is these layers that give the friction couple the expectedevel and stability of coefficient of friction, and wear resistance.t is well known, for instance, that a friction transfer film withthickness of from several to tens microns is likely developedhen organic/sintered pads are set to against grey cast iron.11 Inur knowledge, little study has however been published on theevelopment of friction transfer film or layer on the surface oferamic composites disc, though tribological study implies that

uch a transfer film could be developed under certain conditions.or example, Paris et al.12 noticed that both steel and aluminaeramic pins could generate an adhering layer on the friction

eat

eramic Society 30 (2010) 3187–3201

rack of Cf/SiC composites; they speculated that it was the dif-erence in the development of the layers and their lifecycle thatade the difference in friction behaviour and wear rate.In this paper, we will systematically observe the development

f friction surface in the early stage of braking, which is likelynside the regime of so-called bedding stage of a friction brake.he friction surface will be registered using optical microscopy

o illustrate the evolution of friction surface on the compos-te surface; scanning electron microscopy (SEM) and energyispersive X-ray (EDX) analysis are also used for any furtherharacterisation of the friction surface. The characteristics andossible developing mechanisms of the friction surface will beeneralised and discussed.

. Experimental procedures

.1. Testing components preparation

The Cf/C-SiC composites used in this study was supplied byndustry. The composites were manufactured through chemicalapour infiltration (CVI) to generate carbon fibre/carbon (Cf/C)reforms first, followed by a liquid silicon infiltration (LSI) pro-ess to partly convert the carbon matrix into silicon carbide whilehe silicon melt were filling into channels among Cf/C regions.he as-manufactured Cf/C-SiC composite blocks were cut intoiscs in laboratory with a diameter of 50 mm and a thickness ofbout 15 mm. Two countersink holes with a diameter of 7 mmere drilled on a circular with a diameter of 22 mm on the discs

or bolting the tested discs on the sample jig, which is attachedo the axle, as shown in Fig. 1.

All testing surfaces of the discs were finished following theame procedure: the as-received surfaces of Cf/C-SiC discs wereround on a grinder (Struers Tegrapol-25, Denmark) with dia-ond grinding plates starting from a grit size of about 68 �m

220 mesh), then 26 �m (600 mesh), and 15 �m (1200 mesh),ith a removing thickness of about 300–400 �m per step; theround surfaces were polished with 3 �m diamond paste on aolisher (Struers MD Piano grinding disc, Denmark), to ensureny damages left in previous step were completely removed.

Non-asbestos organic (NAO) brake pads were selected in thistudy. The pads were mounted on a steel back plate in a ringhape with an outside diameter of 48 mm, inner diameter 40 mm,nd thickness 12 mm. The surfaces of the pads were ground ongrinding wheel with SiC grits in a size of 3 mm before used

or the first braking.

.2. Friction tests

The friction tests were conducted on a laboratory-scaleynamometer; the schematic diagram of the assembles of theynamometer is shown in Fig. 1. This dynamometer can pro-ide an inertia of 0.0007 kg m2 and capability of operation withrotation speed up to 18,000 rpm and braking force up to 800 N.

nergy being provided through a spinning flying wheel driven byn air motor. The gas cylinder and solenoid valve are used to con-rol the engagement of the pad with the spinning disc through an

Y. Wang, H. Wu / Journal of the European Ceramic Society 30 (2010) 3187–3201 3189

F (1) axs (11)t asure

otraaipfic

μ

wRai∼it

bdfnblta

atahtto

pdp

2

eotcpntt

ig. 1. Schematic diagram of the assembles of a laboratory-scale dynamometer:peed sensor, (7) sample jigs, (8) samples, (9) torque arm, (10) linear load cell,hermal couple, (16) bearing, (M1) brake load measure, (M2) angular speed me

peration of the brake pedal. The pressure valves are used to con-rol the disc rotation speed and braking pressure. The signals ofotation speed, normal braking load, braking torque and temper-ture inside the pad, are directly collected by a data logger withcapture frequency of 50 Hz; these signals are then converted

nto numerical data by a software. Throughout all the tests in thisaper, a constant disc rotation speed of 11,000 ± 200 rpm and axed normal braking force of 540 ± 20 N were maintained. Theoefficients of friction were calculated using following equation:

= T

W × Re

(1)

here T is the braking torque, W the normal braking load ande is the effective radius (22 mm here) of friction contactingrea. Under the current testing condition, the pad unit pressures about 0.9 MPa, speed at outside diameter (OD) of the pad

25,000 mm/s and at inner diameter ∼21,000 mm/s; these test-ng conditions on this small scale dynamometer are comparableo most of the vehicle testing condition.

The measurement of COF was completed on each discreteraking that led the rotating disc to complete still (here, eachiscrete braking is called one stop). The braking time neededor each stop was about 3 s, and the time interval for the start ofext braking after a stop was set as 6 s. For any new setup with

oth the rotor and pad containing fresh surfaces, an accumu-ated number of 49 stops were accomplished. In order to followhe development of friction surface, observing points were setfter braking stop at the 1st, 2nd, 3rd, 6th, 9th, 14th, 19th, 29th

sebv

le load cell, (2) spacer, (3) stand, (4) flying wheel, (5) gear and belt, (6) angularpressure cylinder, (12) solenoid valve, (13) pressure valve, (14) air motor, (15), (M3) temperature measure, (M4) torque measure.

nd 49th, where the disc and pad were taken off from the fric-ion testing rig, and re-fixed as they were in the original setup,fter completing all examinations. One observation is calledere for all measurements between the closest observing points;hese measurements included COFs of all stops, and microstruc-ure details on the friction surface after the last stop inside anbservation.

To ensure the reliability of the testing data, the same testattern was repeated up to 4 times using the same Cf/C-SiCisc with its surface being re-finished by following the samerocedure after each set of tests.

.3. Friction surface characterization

The friction surface of Cf/C-SiC at each observing point wasxamined and registered by imaging the selected regions withptical microscopy (OM) (MeF3, Reichert-Jung, Wien, Aus-ria). The key phases, e.g. silicon carbide, graphite, silicon andarbon fibre filaments were differentiated by the contrast underolarized lighting condition, as long as the original surface wasot yet completely covered by friction film. When any other fea-ures appeared in OM images, SEM/EDX was used to identifyhe major chemical elements and microstructure.

Secondary-electron imaging and EDX analysis of friction

urfaces were accomplished on a field emission scanninglectron microscope (Leo S360, Cambridge Instruments, Cam-ridge, UK). In most cases, the operating electron accelerationoltage was set to 5 kV; when further information was needed

3190 Y. Wang, H. Wu / Journal of the European Ceramic Society 30 (2010) 3187–3201

Fig. 2. Measurements of coefficient of friction (COF) against the accumulatedbraking stop numbers. Each observation includes the 1st stop after refitting thefriction couple to that where the couple was dismantled from the rig for surfaceeo

ov

3

3

avtsooo

aswcCstewbFtpsuaC

Fig. 3. As-polished surface of Cf/C-SiC disc before braking test. Image in (a)shows two Cf/C regions consisting of in-plane Cf// or needled Cf⊥, large Si andSp

mw

3C

3

Tiofifsi

xamination. Observations I–III include one braking stops; other have more thanne braking stop.

n the variation of constituents underneath top surface, otheroltages could be chosen.

. Experimental results

.1. Coefficient of friction (COF)

The measurements of COF are plotted in Fig. 2 against theccumulated braking stop numbers. For each stop, the averagedalue, with its standard deviation, is a result of four repeatedests using the same disc and pad materials under the sameetup and operating conditions. The results are also grouped intobservations as shown in Fig. 2, and measurements inside eachbservation are linked with a line to show the trend of variationf COF with braking stops.

Inside each observation, the measurements of COF show anscending trend with the increase of the accumulated brakingtops (see observations IV–IX), apart from observation I–IIIhere only one braking stop is included. In any observations

onsisting of multi braking stops, the first stop gives the lowestOF, comparing to the others following it; the COFs of the first

tops in all observations are mostly at a similar level. Amonghese first brakings, the very first one in observation I has the low-st COF, and the one in observation IX gives the highest value,hilst the average values and standard deviations for both firstrakings are inside the ranges of error bars of other first brakings.rom observation IV to IX, after same number of braking stops,

he COF in the later observation is generally higher than that inrevious one. It is also noticed that from observation VII, the

econd braking clearly boosts the level of COF, and then a grad-al improvement is followed for each further braking. Amongll observations, none of these shows that the measurements ofOF have reached a plateau, an indicator that is usually used to

sodi

iC regions between Cf/C regions, and (b) the large cracks across SiC, smallorosity and large voids on the polished surface of SiC.

ark the end of bedding process, even in the last observationhere most stops were applied.

.2. Friction induced evolution of the surface of thef/C-SiC composite

.2.1. Overview of as-polished surfaceThe OM images of an as-polished surface are shown in Fig. 3.

wo types of oriented carbon fibres exist on the surface: nearlyn-plane fibres (Cf//) with fibre lay-out being inside or slightlyff the friction plane, and nearly perpendicular fibres (Cf⊥) withbre directions nearly perpendicular to friction surface. Theormer accounts for the majority of the fibres exposed on theurface, and the later is believed to be the needling fibres for hold-ng the fibre laminae together. Both types of fibre orientation are

hown in Fig. 3(a), as marked. Around each fibre filament, a layerf carbon is surrounded; these carbon regions show bright andark contrast under polarized lighting condition, which mightndicate that the carbon matrix could be a graphite region with

ean C

fcpbc

t

Fbtr

Y. Wang, H. Wu / Journal of the Europ

airly uniform structure (n.b. we will hereafter call these regions

arbon, carbon matrix or C-matrix with no implication of anyarticular crystallography of carbon). Among the fibre bundlesonded and embraced by the carbon matrix, there exist siliconarbide and silicon regions, which show grey and white con-

saai

ig. 4. Registered OM images to illustrate the development of friction surface in silraking test; (b), (c), (d) and (e) the surface after the 2nd, 3rd, 19th and 49th brakinype C friction surface that consists of friction transferred layer on top and displays aespectively. (For interpretation of the references to colour in this figure legend, the r

eramic Society 30 (2010) 3187–3201 3191

rast, respectively. In microstructural aspect, voids with a size

cale up to tens microns and large cracks/crevices exist on thes-finished surface, as well as small porosity within a scale offew microns, inside both SiC and carbon regions, as shown

n Fig. 3(b). In summary, the friction surface is composed of

icon regions of the composite: (a) as-finished surface in silicon regions beforeg stop, respectively. The arrowed areas are categorized as type A, type B andcontrast of brownish, white, and dark grey or a mixture of dark grey and white,eader is referred to the web version of the article.)

3 ean Ceramic Society 30 (2010) 3187–3201

aa

3

oooUtlamfBpaifbB

timsFctbIomcalfobewtttAv

3

iihavctA

Fig. 5. Secondary-electron images of silicon regions where three types of fric-tion surface appeared in OM imaged (see Fig. 4(b)) when the composite discwas subjected to two braking stops: (a) region with brownish contrast (type A);(b) regions with white contrast (type B); (c) region with a mixture of grey andwhite colour (type C). Inset in each image is the EDX spectrum obtained fromthis region under an accelerating voltage of 3 kV. (For interpretation of the ref-

192 Y. Wang, H. Wu / Journal of the Europ

structure with a clear hierarchy arrangement of multi phases,long with voids/porosity, large and micro-cracks.

.2.2. Surface evolution in silicon regionsA series of registered OM images in Fig. 4 show the devel-

pment of friction surface in a silicon region. With the increasef accumulated numbers of braking stop, a friction transfer filmr layer has been built up continuously on the top of this region.nder the current imaging condition, the friction surface after

wo stops shows three types of contrast: type A in brownish (oright grey in black and white printing), type B in bright white,nd type C a mixture of dark grey and white or dark grey only, asarked in Fig. 4(b); it is evident that type A film prevails on the

riction surface, whilst type B and C exist in isolated patches.y increasing one more braking stop, more type B and type Catches appears, with a significant shrinkage of type A area, asrrowed in Fig. 4(c). Up to 19 stops, type A contrast is observedn scattered small areas only, but type B and C predominate theriction surface (Fig. 4(d)). By further increasing the number ofraking stop to 49, the friction surface shows predominant typeand C contrasts (Fig. 4(e)).In order to examine the chemical composition of the fric-

ion surface where three different contrasts appear under OMmaging condition, low-voltage SEM-EDX analysis was imple-

ented on the surfaces. SEM images and corresponding EDXpectra from areas with different contrasts are summarised inig. 5. The friction surface with type A contrast is copper-rich,omparing to other areas; type B is iron rich, and type C con-ains both copper and iron in a significant amount, as indicatedy the intensity of the L lines of Fe and Cu in the EDX spectra.nside the type A and type C, there exists a significant amountf magnesium and oxygen, while in type B, both oxygen andagnesium are in a very low level, if any. It is yet impossible to

onfirm whether the existence of oxygen is related to the appear-nce of other elements, e.g. copper/magnesium, at the currentevel of characterization. The morphology of the friction sur-ace (actually the friction transfer film or layer, as the chemicalsn the surface are different from the composites already) cane differentiated among these three types from the secondary-lectron images: both the type A and C look a bit of crumblingith visible granulated materials in the film, though the size of

he crumbles in the type A is much smaller than in the type C;he type C has a rougher surface, comparing to the type A, withhe existence of craters, voids and cracks. In contrast to the type

and C, the type B film has a smoother surface with much fewerisible particles and voids.

.2.3. Surface development in silicon carbide regionsThe friction surface development in silicon carbide region

s shown in Fig. 6. Under an imaging condition of differentialnterference contrast (DIC), a number of small voids/craters areighlighted on the polished surface of this region (see Fig. 6(a)nd Fig. 2). Right after the first braking stop, some of these

oids/craters are filled with transferred materials showing darkontrast, as shown in the arrowed sites, and a few in bright con-rast; they are marked as type B and C respectively in Fig. 6.fter 6 braking stops, these type B patches grow up to a larger

erences to colour in this figure legend, the reader is referred to the web versionof the article.)

ean C

ealiei

Nf

Faod

Y. Wang, H. Wu / Journal of the Europ

xtend, and small parts of the type B convert into type C alike,s shown in Fig. 6(c). Subjected further braking stops, these

arge white patches, along with the small white spots appearn the middle area in Fig. 6(c), are partly or completely cov-red with type C contrast layers, as seen in the surface imagesn Fig. 6(d), (e) and (f) after 14, 19 and 49 stops, respectively.

bas

ig. 6. Registered OM images on the friction surface of a silicon carbide region: (a)fter the 1st, 6th, 14th, 19th and 49th braking stop, respectively. The arrows indicatef friction surface showing similar contrast as type B and type C friction surface appeamage followed by a deposit of transferred materials on SiC surface.

eramic Society 30 (2010) 3187–3201 3193

ote, no type A contrast area is observed on this friction sur-ace.

It is noticed that after the first stop, the friction surface near theoundary of the silicon carbide region shows type C contrast andcrumbling appearance. When braking stop number increases,

uch kind of friction surface develops continuously toward the

as-finished surface before braking; (b), (c), (d), (e) and (f) the friction surfaceplaces where transferred materials were deposited; B and C represent the typeared in Fig. 4; D is an area highlighted to notice the journey of the mechanical

3 ean C

csFsstsc

DlrdmucCh

simpdtcmmtittfitismt

3

satctAamCtiwfocr

Btb

3i

rfmafcmaiScafis

3

ftikaicaAi

4

Cosaoasmiaaitf

194 Y. Wang, H. Wu / Journal of the Europ

entral area of the region. After 49 stops, most of the frictionurface in the SiC region shows type C contrast, as shown inig. 6(f). It is also noted that before type C appears on the frictionurface, extensive cracking has occurred on the SiC surface;ome surface area tends to chip off due to lateral cracking. Up tohe braking numbers of current test, only the core of this regioneems to be in good integrity without clear change of imageontrast.

It is worth noting that the development of a circulated regionin the upper part of this area. Images from (a) to (f) show a

ateral cracking has been developed underneath the surface inegion D between the 1st and 6th stop, and then the partially-etached surface is crushed into many pieces with transferredaterials filling in the cracks between the 6th and 14th stop;

p to the 19th stop, the whole area is converted into a regionompletely filled with transferred materials, with a typical typecontrast; to the 49th stop, the area filled with transfer material

as expanded further (see Fig. 6(f)).This SiC region was further analyzed with SEM/EDX after 49

tops (Fig. 7). The EDX measurements confirm that the regions covered or filled with transferred materials. The transferred

aterials showing type C contrast are composed of iron, cup-er, magnesium and oxygen; the region with white contrast isominated with iron, as shown by the EDX spectrum at posi-ion a and c respectively in Fig. 7(a), which should be a furtheronfirmation of the transfer layer belonging to the type B. In theiddle area, there is no evidence to show that any transferredaterial has been deposited on the surface of silicon carbide, as

he EDX spectrum shows no more than carbon and silicon linesn position b. However, in another region, EDX spectrum showshat transfer film has deposited on the surface of silicon carbide,hough very thin, as shown in Fig. 7(c). Under a higher magni-cation on the damaged region with type C contrast (Fig. 7(b)),

he transferred materials are exposed, and clearly they all fillnside surface cracks; the surface region looks like granulatedilicon carbides, with a size ranging from sub-micron to a fewicrons, that are held together by the transferred materials on

he surface.

.2.4. Cf/C region containing vertical carbon fibres (Cf⊥)After first braking stop, the carbon matrix is subjected to

ignificant mechanical damage, and simultaneously the damagereas are filled with transferred materials or/and crumbling par-icles; same situation occurs on the interfaces between verticalarbon fibres and carbon matrix, as shown in Fig. 8(b). However,he vertical carbon fibres themselves are still in good position.fter 49 brakings, majority of the region, including carbon fibre

nd carbon matrix, has been damaged, and left with transferredaterials, as shown in Fig. 8(c). There is no clear type B andcontrast that could be differentiated under the imaging condi-

ion. Further SEM/EDX analysing results, as shown in Fig. 8(d),ndicate that the voids formed by surface cracking are filled withear debris with chemical constituents that are typical for trans-

erred film with type C contrast observed on the friction surfacesf silicon and silicon carbide. On the cross-sectional surface ofarbon fibres, the smooth and flat part has no transferred mate-ials deposited, as confirmed by the EDX spectrum at position

egfC

eramic Society 30 (2010) 3187–3201

, so does the exposed carbon matrix surface. It is also noticedhat some transferred materials also fill in the damaged interfaceetween carbon matrix and carbon fibre.

.2.5. Surface development in Cf/C regions containingn-plane carbon fibres

Like the vertical carbon fibres, all in-plane fibres are sur-ounded with carbon matrix (Fig. 9(a)). After initial braking,racture occurred in both carbon fibres and carbon matrix, asarked in Fig. 9(b); some fractured parts have been dislodged

nd stripped away, and the left positions are filled with trans-erred materials. When more brakings are applied, the morearbon fibres and matrix are dislodged through fracturing, theore are transferred materials deposited on the friction surface,

s shown Fig. 9(f). It is noted that all the transferred materialsn Cf//C region look fairly crumbling, lack of enough cohesion.EM analysis shows that the transferred materials have the samehemical composition as that with the type C contrast in siliconnd silicon carbide regions; on the exposed surface of carbonbres and matrix, there is no transferred material deposited, ashown in Fig. 9(f).

.2.6. Surface development in large voids and cracks/gapsFig. 10(a) shows a typical large voids on the as-polished sur-

ace. After one braking stop, the void is completely filled withransferred materials; they look fairly dense. After further brak-ng, some of the filled materials are pulled out. As the braking isept going, it seems such regions undergo a process of filling-innd partially-dislodging. As seen in Fig. 10(d), the same voids fully filled with transfer materials again. Fig. 11 shows thatracks in silicon carbide regions on the as-prepared surface arelso filled by transferred materials fairly swiftly during braking.

similar dynamic process of filling-in and partially-dislodgings observed as that appeared in other large voids.

. Discussion

The development of the friction surface on the surface off/C-SiC composite is systematically examined through a seriesf microscopic image registrations on microstructurally repre-entative regions where friction braking is initially executed onn as-machined state by using an organic pad. The knowledgen what may happen on the surface during braking could givemore inclusive understanding of the development of friction

urface through physically interacting between the transferredaterials produced by the friction operation and the compos-

te that is likely composed of different ceramic constituentsnd microstructural characteristics. Such an awareness couldlso light up possibilities of enhancing the overall engineer-ng performance of brakes that consist of ceramic compositeshrough optimising the microstructure of the composites, sur-ace engineering, and fine-tuning of pad materials. By using the

xperimental results, we will give further analysis and possibleeneralisation on the circumstance of friction surface, governingactors on its development, and possible correlations betweenOF and friction surface.

Y. Wang, H. Wu / Journal of the European Ceramic Society 30 (2010) 3187–3201 3195

Fig. 7. Secondary-electron images of the friction surface in silicon carbide region after 49 braking stops: (a) a region corresponding to Fig. 6(f) imaged by OM, andEDX spectra in position a, b and c; (b) higher magnification of a region inside a square with dashed line in (a); (c) another silicon carbide region where EDX spectraat position d and e indicate that transferred materials have been successfully deposited. Accelerating voltage was 3 kV for imaging and EDX spectra acquisition.

3196 Y. Wang, H. Wu / Journal of the European Ceramic Society 30 (2010) 3187–3201

Fig. 8. Registered OM images of the friction surface in a Cf⊥/C region of the composite: (a) as-finished surface before braking; (b) and (c) the surface after the 2nd,and 49th braking stop, respectively. The arrows indicate a position where circular groove was generated along the interface between Cf⊥ and carbon matrix aftertwo braking stops. (d) The secondary-electron image of the friction surface in Cf⊥/C region after 49 braking stops; wear debris accumulated in voids created byfracturing and dislodging, and split interfaces. Spectra from positions A and B indicate no transferred materials being deposited on the surface of carbon matrix andthe cross-section surface of carbon fibre; spectrum from position C indicates that deposited debris in voids/gaps contain Cu, Fe, Mg and O.

ean C

4d

rb

pr

FSo

Y. Wang, H. Wu / Journal of the Europ

.1. The morphology of the friction surface and itsevelopment

Registered microscopic images on post-braking surfaces haveevealed that transferred materials, or sometimes called thirdodies,13 are introduced onto the surface of the ceramic com-

aoto

ig. 9. OM images of friction surface in Cf///C region: (a) as-finished surface beforeecondary-electron image and EDX spectra are presented in (f) for the friction surfaf carbon fibre and most of the carbon matrix.

eramic Society 30 (2010) 3187–3201 3197

osites by the braking operations, when a pad, composed ofesin bonded materials such as metals and minerals, is used to

gainst the ceramic disc. These third body materials are observedn the friction surface right after the initial braking, regardlesshe constituents and the microstructure features on the surfacef the composite, as shown in Figs. 4(b), 6(b), 8(b), 9(b), 10(b)

braking, (b) after 1 stop, (c) after 2 stops, (d) after 3 stops, (e) after 49 stops.ce after 49 braking stops; no transferred materials are identified on the surface

3198 Y. Wang, H. Wu / Journal of the European Ceramic Society 30 (2010) 3187–3201

d befo

amtoFeabmitctp

mtwlaivfhf

Fig. 10. OM images the friction surface in voids: (a) as-finishe

nd 11(b). However, their embedding configuration and existingorphology vary greatly on the surface. On the top of the silicon,

he deposition of the transferred materials could be on anywheref the as-finished surface spreading across the whole region.or silicon carbide, the transferred ones tend to embed in wher-ver that is topographically lower, e.g. micro-cracks, pits, voidsnd porosity; these microstructural feature are likely generatedy the Hertizan contacting damage of abrasives during surfaceachining and friction braking, or borne inside the compos-

tes themselves. In the Cf/C region, the depositing tendency of

he transferred materials is very similar to that observed on sili-on carbide. Certain embedding places are nevertheless believedo be created immediately after the friction contact through aossible process of fracturing first, and then dislodging of the

acta

Fig. 11. OM image of friction surface on cracks: (a) as-finished surface before br

re braking, (b) after 1 stop, (c) after 3 stops, (d) after 49 stops.

aterials; these sites can be distinguished by comparing the fric-ion surfaces after 1–3 braking stops (Figs. 8(b) and 9(b)–(d))ith the as-finished ones (Figs. 8(a) and 9(a)). For pre-existing

arge voids and cracks/gaps, despite their existing locations, theyre all filled with transferred materials immediately after thenitial brakings. It is believed that the structural features of theoids/cracks/gaps on the surface could provide these regions aunction just like a cheese grater does, and the cavities mightave collected most of the filled materials that were chipped offrom the pad, as seen in Figs. 10 and 11. Note, judging from the

ppearances of the compacts in Figs. 10 and 11, their density,ohesion and their longevity may vary a lot with the nature ofhe cavities/gaps, i.e. their depths, in-plane dimensions, shapesnd materials surrounding them. We believe that a further study

aking, (b) after 1 stop, (c) after 3 stops, (d) after 9 stops, (e) after 49 stops.

ean C

wmpoi

atdhts

(

(

(

aststt

4d

ccitlcsotteawtopw

bcpblhsftotgcft

otisaautssatmbfdpb(

fha

G

whm

l

KtF

Y. Wang, H. Wu / Journal of the Europ

ould be required on the compacting behaviour of the filledaterials if any filling-ups of physically low sites is to be pur-

osely adapted to modify the friction/engineering performancef the composites, whilst some industry successes are claimedn this aspect.

After the preliminary deposition of the transferred materi-ls and any further development in the following up braking,he morphology of the friction surface has subjected to a fairlyissimilar route in different regions; details on the developmentave been described in last section. Based on the observations,hree types of friction surface can be generalized in the earlytage of braking, and we categorize them as following:

a) type I, the transferred materials have a relatively stable rela-tionship with the matrix, and any cumulative developmentshould involve an interaction with the previously depositedmaterials and be able to maintain their longevity in a frictiontransfer layer;

b) type II, the transferred materials have unstable relationshipwith the matrix, and the longevity of the friction transferlayer is likely dependant on the durability of the interfacebetween the transferred materials and the matrix, as well asthe sturdiness of the near-surface region of the matrix;

c) type III, the transferred materials have a relatively stablerelationship with the matrix, but the longevity of the frictiontransferred materials is likely determined by the cohesion ofthe compact.

The registered images have revealed that transferred materialsre able to successfully deposit on silicon regions and the frictionurface consisting of the developed layer should belong to theype I; friction surfaces with those transferred materials on theilicon carbide and on the Cf/C regions can be categorised intohe type II; the trapped materials in large voids/gaps existing onhe surface should fit in the type III.

.2. Main factors that could govern friction surfaceevelopment

In the silicon region, it is noticed that the transferred materialsan deposit on the surface straight as long as the pad has a directontact with the surface of the composite; transferred materialsntroduced by the following up brakings are likely fused withhe previous ones to generate a so-called friction transferredayer over the substrate. Judging from the observed variation inontrast of the OM images on the post-braking surfaces insideilicon regions (Fig. 4), together with the corresponding resultsf chemical constituent analysis (Fig. 5), we believe that a rela-ively stable transfer layer might have been built up layer by layerhrough a fusion of metallic materials and a dispersion of min-rals inside them during each braking operation, following theppearance of an initial metallic layer right on the top of silicon,hich seems to be strongly bonded. This inference is similar

o the observed development of transfer layer on the surfacesf friction couples consisted of grey cast iron rotor and organicad.14,15 Further details on the bonding of the initial materialsith the silicon, the microstructure of the transfer layer and its

rsct

eramic Society 30 (2010) 3187–3201 3199

uilding up procedure can be identified and reconstructed withross-sectional TEM analysis, but this is out the scope of thisaper. It is worth mentioning, however, that two factors mighte the essential for the successful development of type I transferayer: first, the initial deposition of transferred materials mustave a stable relationship with the substrate; secondly, the sub-trate must be able to keep its integrity during braking. Theormer one could be realized through a strong bond between theransferred materials and the substrate; in these silicon regions,ur recent analysis has shown one of the bonding possibilities ishrough a fusion of cupper and silicon under high temperaturesenerated by the friction.16 The later one is confirmed by theurrent observation that in the braking testing regime, no visibleracture damage of the silicon substrate has yet been seen up tohe last braking.

In silicon carbide and carbon fibre/carbon regions, the devel-pment of transfer layer shows a tendency to be influenced byhe unsteadiness of the substrate, if not the lack of strong bond-ng or adhesion only between the transferred materials and theurface of carbon and silicon carbide. Surfaces in both regionsre subjected to breaking damage from the very early brakings,s shown in Figs. 7–9; such damage could obviously lead to annstable subsurface region. The argument could be that if theransferred materials had had a better adhesion with the sub-trates and/or the substrates had a higher fracture resistance, theurface of the substrates could have suffered less contact dam-ge and then been able to hold up the transfer layers. Underhe current observations, no evidence shows that the transferred

aterials have a strong adhesion with silicon carbide and car-on, whereas a firm support on the inference can only be done byurther TEM analysis. The appearance of significant contactingamage in both silicon carbide and carbon regions implies oneossibility that the critical energy releasing rates of silicon car-ide (GC

SiC) and carbon (GCC) could be lower than that of silicon

GCSi) in this composite, i.e. GC

SiC < GCSi and GC

C < GCSi.

Fracture releasing rate, GC, is a parameter used to gauge theracture tolerance of a material under loading condition. It shouldave the following relationship with the Young’s modulus, E,nd fracture toughness, KIC, of a material

C = (1 − ν2)K2IC

E(2)

here ν is its Poison ratio. By re-arranging Eq. (2), we shouldave the following linear relationship between logarithm ofodulus and fracture toughness.

og10 KIC = 1

2log10

(E

1 − ν2

)+ 1

2log10 GC (3)

In a log10(KIC)–log10(E/1 − ν2) chart, the intercepts on theIC axis of parallel linear lines with a slope of 1/2 represent

he values of GC. A group of such lines have been drawn inig. 12, with specific values are labelled for the fracture strain

ates. When materials sit on the same line, it indicates that theyhould have same fracture strain rate; materials near the top leftorner should have larger fracture strain rate than those closero the bottom right corner. More details on using this chart are

3200 Y. Wang, H. Wu / Journal of the European C

Fig. 12. Fracture toughness of ceramics (KIC) plotted against Young’s modulus(E, or E/(1 − ν2)). The family of lines are of critical energy releasing rate (GC).Thc

aScoSTailtrapditctcmtcs

ustopi

imntlbegsnodwcorsbsgiiciotamcodnt

4

tmguisttsitrstlr

he shaded band shows that pure graphite, silicon and isotropic carbon fibresave same fracture energy; any ceramics above this band are more resistant toracking than those on/below the band.

vailable from Ref. 17. Data drawn from Cambridge Materialselection database18 have been marked in this chart for theseeramics that could be comparable with the ceramic constituentsbserved in this composite: silicon, sintered SiC, nitride bondediC, vapour deposited SiC, pure graphite, isotropic carbon fibre.hree carbon fibre reinforced carbon composites (CCC1, CCC2nd CCC3) have also been included.19 As shown in Fig. 12, sil-con and graphite have the similar fracture strain rate, but muchower than any SiC ceramics and Cf/C composites. This implieshat cracking damage would not have occurred in SiC and Cf/Cegions when no cracking is observed in silicon. The discrep-ncy between the experimentally observed and theoreticallyredicted on the fracture resistance of the ceramic constituentsuring braking may imply that KIC values measured from ceram-cs might have over-estimated, when these data are intendedo be applied to measure the micro-cracking resistance of theonstituents on the friction surface during braking. In fact, frac-ure resistance of polycrystalline ceramics could increase withrack extension, i.e. the well-known R-curve behaviour, and theagnitude of this increase could be in excess of a factor of

hree.20 For ceramics with complex microstructure, like poly-rystalline and composites, this difference would become moreignificant.21

Nevertheless, we think Fig. 12 could provide a guideline fornderstanding, designing and engineering these ceramic con-tituents for better friction performance of the composites, if

he KIC values are measured to genuinely gauge the resistancef microcracking or cracking initiation, rather than large crackropagation. Now, we believe that the following possibilitiesnside the silicon carbide might have led its surface more frag-

tTff

eramic Society 30 (2010) 3187–3201

le than silicon: (a) the silicon carbide regions in the compositeight have been weakly bonded as the grain boundaries were

ot purposely engineered in the LSI process, which could leado a degradation of fracture toughness at the microstructureevel, which was shown a possibility of less than 2.1 MPa m1/2

y Padture and Lawn22; (b) large residual tensile stress couldxist on the grains boundaries due to the constraint on therowth of silicon carbide grains in LSI process and the thermaltrains generated during cooling down from the silicon impreg-ation temperature, which are featured through the appearancef stacking fault inside the SiC grains.23,24 This view has beenemonstrated in a study on the wear resistance of alumina,here the damage resistance has little relationship with large

rack propagation resistance, i.e. KIC measured by most meth-ds, but a strong relationship with short crack energy releasingate.25 The same analysis can apply to Cf/C region where thehort-crack energy releasing rate, particularly along the interfaceetween carbon fibre and carbon matrix, should be significantlymaller than those measured from Cf/C composites or typicalraphite, since in the design of most fibre reinforced compos-tes, the fibre/matrix interfaces are deliberately weakened tomproved the fracture tolerance,26 which is why the three CCComposites, shown in Fig. 12, are sitting above all SiC ceram-cs. Unfortunately little experimental measurement is availablen the interface or grain boundary fracture energy, as an addi-ion in Fig. 12 to exhibit the picture more clearly. However, were convinced through the analysis that short crack initiating oricro-crack propagating resistance must be improved for the

eramic matrix, including all interfaces in these composites inrder to promote the development of a steady friction surfaceuring braking for the aim of maintaining the friction and engi-eering performance of the composites, in addition to improvehe adhesion of the third body materials on the friction surface.

.3. Coefficient of friction and friction surface development

The development of these three types of friction surfaces onhe surface of the composites could make a direct impact on the

easurements of coefficient of friction and its stabilization. Therowth of type I likely leads to a stable friction, and type II to anstable friction; type III hardly gives an unstable friction, butts amount on the examined surface may be too small to give anyignificant influence in this study. In general, the growth of theransferred layer with the addition of further brakings increaseshe contact area between the disc and the pad; such an incrementhould lead to an improvement of the measured COFs, as shownn Fig. 2. Once the development of the transfer layer is stabilized,he COF would reach to a constant value, where it is generallyegarded as the completion of bedding process. However, thistabilization has not yet been seen in Fig. 2 under the currentesting regime. Judging from the development of the transferayer on different regions, we believe that those on the siliconegions could give a stabilized state very swiftly, but much longer

ime is needed for those on the silicon carbide and Cf/C regions.herefore, this unstable type II transfer layer could be the key

actor that leaves Cf/C-SiC composite longer bedded-in time in ariction brake. In the tested composite, majority of the surface is

ean C

cseo

5

fiocdtscaIstsaac

trsastfarIst

A

TYt

R25. Cho SJ, Hockey BJ, Lawn BR, Bennison SJ. Grain-size and R-curve effects

Y. Wang, H. Wu / Journal of the Europ

omposed of Cf/C and SiC, and Si is less than 10%; the frictionurface is hence dominated with type II friction surface. Wenvisage it would take longer time to achieve a stable coefficientf friction.

. Summary

The development of friction transfer layer is studied on theriction surface of Cf/C-SiC composite when the composite discs coupled with an organic pad. The transfer layer is composedf third body materials containing iron and copper as the mainhemical elements; its development during friction stronglyepends on the adhesion between the transferred materials andhe ceramic matrix, as well as the steadiness of the ceramicubstrate. Three types of friction surface are identified on theomposite’s surface: type I – the friction transferred layer hassteady relationship with the matrix and good longevity; type

I – that the friction transferred layer has an unstable relation-hip with the matrix and poor durability; type III – the frictionransferred layer has a steady relationship with the matrix buthort lifetime. It appears that type I exists on all silicon regionsnd some silicon carbide regions; type II on all Cf/C regionsnd most of silicon carbide regions; type III in large voids andracks/crevices.

The measured coefficients of friction show an increasingrend with the accumulation of braking stops, but did not yeteach to a stabilized level in the current test regime. These mea-urements echo the fact of friction surface development: theppearance of type I likely leads to a steady increase in COFwiftly, but the type II to an unstable friction and longer brakingime to approach a stable level. One of the controlling factorsor type II layer should be the fracture resistance of the matrix,part from the adhesion strength between the transferred mate-ials and the ceramic matrix. To avoid the appearance of typeI surface, it is necessary to increase the critical strain rate ofhort crack propagation of silicon carbide and Cf/C regions inhe composites.

cknowledgements

The authors would like to acknowledge the EPSRC and theechnology Strategy Board in UK for providing studentship forW. Mr. J. Bates and Dr. D. Ross are acknowledged for their

echnical support on using FEGSEM and MEF3 microscopy.

eferences

1. Krenkel W, Heidenreich B, Renz R. C/C-SiC composites for advancedfriction system. Adv Eng Mater 2002;4(7):427–36.

2. Fan S, Zhang L, Xu Y, Cheng L, Lou J, Zhang J, et al. Microstructure andproperties of 3D needle-punched carbon/silicon carbide brake materials.Compos Sci Eng 2007;67(11–12):2390–8.

eramic Society 30 (2010) 3187–3201 3201

3. Xu Y, Zhang Y, Cheng L, Zhang L, Lou J, Zhang J. Preparation and frictionbehaviour of carbon fibre reinforced silicon carbide matrix composites.Ceram Int 2007;33(3):439–45.

4. Krenkel W. Carbon fibre reinforced CMC for high-performance structures.Int J Appl Ceram Technol 2004;1(2):188–200.

5. Wang J, Jin Z, Qiao G. Rapid fabrication of C/C-SiC composites. Key EngMater 2006;317–318:159–62.

6. Cai YC, Xu Y, Li B, Fan S, Zhang L, Cheng L, et al. Low-cost prepara-tion and frictional behavious of a three-dimensional needled carbon/siliconcarbide composites. J Eur Ceram Soc 2009;29(3):497–503.

7. Zhang Y, Xu Y, Lou J, Zhang L, Cheng L. Braking behaviour of C/SiCcomposites prepared by chemical vapour infiltration. Int J Appl CeramTechnol 2005;2(2):114–21.

8. Jiang G, Yang J, Xu Y, Gao J, Zhang J, Zhang L, et al. Effectof graphitization on microstructure and tribology properties of C/SiCcomposites prepared by reactive melt infiltration. Compos Sci Technol2008;68(12):2468–73.

9. Krenkel W, Berndt F. C/C-SiC composites for space applications andadvanced friction system. Mater Sci Eng 2005;A412(1–2):177–81.

10. Stadler Z, Krnel K, Kosmac T. Friction behaviour of sintered metal-lic brake pads on a C/C-SiC composite brake disc. J Eur Ceram Soc2007;27(2–3):1411–7.

11. Liu T, Rhee SK, Lawson KL. A study of wear rates and transfer films offriction materials. Wear 1980;60(1):1–12.

12. Paris JY, Vincent L, Denape J. High-speed tribological behavioursof carbon/silicon-carbide composite. Compos Sci Technol2001;61(3):417–23.

13. Godet M, Play D, Berthe D. An attempt to provide a uniform theory oftribology through load-carrying capacity, transport and continuum mechan-ics. J Lubr Technol 1980;102(2):153–64.

14. Österle W, Griepentrog M, Gross T, Urban I. Chemical and microstructuralchanges induced by friction and wear of brakes. Wear 2001;251:1469–76.

15. Österle W, Urban I. Third body formation on brake pads and rotors. TribolInt 2006;39:401–8.

16. Wang Y, Wu H, unpublished work.17. Ashby MF. Materials selection in mechanical design. 3rd ed. Oxford:

Elsevier; 2008. pp. 45–78.18. Cambridge Engineering Selector, Version 5.1.0. Granta Design Ltd.; 2009.19. Lackey WJ. In: Buschow KHL, Cahn R, et al., editors. Encyclopaedia of

materials: science & technology. Elsevier Science Ltd.; 2001. p. 952–67.20. Cook RF, Lawn BR, Fairbanks CJ. Microstructure-strength properties

in ceramics. I. Effect of crack size on toughness. J Am Ceram Soc1985;68(11):604–15.

21. Bennison SJ, Lawn BR. Role of interfacial grain-bridging sliding friction inthe crack-resistance and strength properties of nontransforming ceramics.Acta Metal 1989;37(10):2659–71.

22. Padture NP, Lawn BR. Toughness properties of a silicon carbide withan in situ induced heterogeneous grain structure. J Am Ceram Soc1994;77(10):2518–22.

23. Lawn BR. Fracture of brittle solids. 2nd ed. Cambridge, U.K: CambridgeUniversity Press; 1993. pp. 319–24.

24. Leatherborrow A, Wu H. The nature of silicon carbide phases developedfrom different carbonaceous sources and its impact on the microstructureof Cf/C-SiC composites. In: Proc 34th Int Conf & Exp on Adv Ceram EngSci Proc 31 (8), in press.

in the abrasive wear of alumina. J Am Ceram Soc 1989;72(7):1249–52.26. Hull D, Clyne TW. An introduction to composite materials. In: Cambridge

solid state science series. 2nd ed. Cambridge: Cambridge University Press;1996. pp. 147–53.